When AI is asked to give short, precise answers, it can lead to higher error rates, according to people who run an AI testing platform.
AI & Journalism Links
This started out as a spreadsheet. Now it's a blog. And a Newsletter.
The New Yorker is doing a really great job right now of pushing the conversation about AI and knowledge professions in a smart, forward-thinking direction. Latest example: An article about the humanities by D. Graham Burnett.
Ezra Eemans’ keynote at the Nordic AI in Media Summit: How news organizations can adapt and compete with “unlimited” information sources. (LinkedIn)
Video: AI prompt engineering deep dive (Anthropic)
How do newsrooms tackle the challenges of generative AI? Hands-on use cases and Q&As in the EBU News Report 2025.
Florent Daudens’ appeal at Nordic AI in Media Summit: Newsrooms must “get real” and build for their fickle, platform-native audience. (LinkedIn)
GenAI in media – 90+ tools and products from around the world. (Kalle Pirhonen, Numeroiden takaa)
Data analysis challenges Google’s AI Overviews claims, suggesting 34.5% fewer clicks.
“They may feel that AI has come for them, and so their reaction (and therefore their language) is visceral and raw“, writes David Caswell about how some journalists are getting defensive and what conversation we need now. (Reuters Institute)
Reuters Institute’s highlights from the International Journalism Festival 2025 in Perugia.
Transformer Lab is an open-source platform for building, tuning, and running LLMs locally, sans coding.
Seamless MCP-powered integrations sound appealing, but raise performance and security concerns. (Shrivu’s Substack)
Deep-dive into RAG and evaluation: How Süddeutsche built their election chatbot. (Medium)
Asking the tough question: “Why do AI company logos look like buttholes?” (Radek Sienkiewicz, VelvetShark)
“Stories that just could not be told without the assistance of AI”: NYT’s Zach Seward leans on AI tools, yet remains skeptical of AI-generated content. (Depth Perception)
